动手深度学习-硬件对于训练的影响
CPU训练
在计算a+b的数据需要准备a和b
- l1访问延迟是0.5ns
- l2访问延迟是7ns
- 主内存访问延迟100ns(200xl1)
提升空间和时间的内存本地性
- 时间：重用数据使得保持他们在缓存中
- 空间：按需读写数据可以使得预读取
所以一个矩阵如果按列存储，访问一行会比访问一列要快
GPU
每个绿点是一个小的处理器(这是泰坦x)
cpu和核心数在6/64-2k/4k
每秒计算的浮点数就是核的数量乘以主频。
内存带宽显卡也是十倍甚至上百倍于GPU，cpu为30GB-100GB，而GPU为400GB-1TB
但是CPU控制流强于GPU
如何提升GPU利用率
并行：使用数千个线程（也叫流处理器）
内存本地访问性：哈运存更小，架构耿凯但你
少用控制语句：支持有限，同步开销很大
不要频繁在CPU和GPU之间传输数据
CPU高性能使用C++，编译器成熟
GPU NVIDIA上用CUDA，OpenCL，质量取决于硬件厂商
单机多卡并行
一台机器能装1-16个GPU在训练和预测时，将一个小批量计算切分到多个GPU上来达到加速的目的
常用的切分方案有
- 数据并行：将小批量分成n块，每个GPU拿到完整的参数计算这一块数据的梯度，性能通常更好。
- 模型并行：将模型分成n块，每个GPU拿到一块模型计算它的前向结果，通常用于模型大道单GPU放不下
- 通道并行（数据+模型并行）
需要模型并行的场景是一张卡放不下模型。
多GPU代码实现
从0开始实现多GPU训练
import torch
from torch import nn
from torch.nn import functional as F
from d2l import torch as d2l
scale = 0.01
W1 = torch.randn(size=(20, 1, 3, 3)) * scale
b1 = torch.zeros(20)
W2 = torch.randn(size=(50, 20, 5, 5)) * scale
b2 = torch.zeros(50)
W3 = torch.randn(size=(800, 128)) * scale
b3 = torch.zeros(128)
W4 = torch.randn(size=(128, 10)) * scale
b4 = torch.zeros(10)
params = [W1, b1, W2, b2, W3, b3, W4, b4]
# 定义模型
def lenet(X, params):
h1_conv = F.conv2d(input=X, weight=params[0], bias=params[1])
h1_activation = F.relu(h1_conv)
h1 = F.avg_pool2d(input=h1_activation, kernel_size=(2, 2), stride=(2, 2))
h2_conv = F.conv2d(input=h1, weight=params[2], bias=params[3])
h2_activation = F.relu(h2_conv)
h2 = F.avg_pool2d(input=h2_activation, kernel_size=(2, 2), stride=(2, 2))
h2 = h2.reshape(h2.shape[0], -1)
h3_linear = torch.mm(h2, params[4]) + params[5]
h3 = F.relu(h3_linear)
y_hat = torch.mm(h3, params[6]) + params[7]
return y_hat
# 交叉熵损失函数
loss = nn.CrossEntropyLoss(reduction='none')
def get_params(params, device):
new_params = [p.to(device) for p in params]
for p in new_params:
p.requires_grad_()
return new_params
new_params = get_params(params, d2l.try_gpu(0))
print('b1 权重:', new_params[1])
print('b1 梯度:', new_params[1].grad)
def allreduce(data):
for i in range(1, len(data)):
data[0][:] += data[i].to(data[0].device)
for i in range(1, len(data)):
data[i][:] = data[0].to(data[i].device)
data = [torch.ones((1, 2), device=d2l.try_gpu(i)) * (i + 1) for i in range(2)]
print('allreduce之前：\n', data[0], '\n', data[1])
allreduce(data)
print('allreduce之后：\n', data[0], '\n', data[1])
data = torch.arange(20).reshape(4, 5)
devices = [torch.device('cuda:0')]
split = nn.parallel.scatter(data, devices)
print('input :', data)
print('load into', devices)
print('output:', split)
def split_batch(X, y, devices):
"""将X和y拆分到多个设备上"""
assert X.shape[0] == y.shape[0]
return (nn.parallel.scatter(X, devices),
nn.parallel.scatter(y, devices))
def train_batch(X, y, device_params, devices, lr):
X_shards, y_shards = split_batch(X, y, devices)
# 在每个GPU上分别计算损失
ls = [loss(lenet(X_shard, device_W), y_shard).sum()
for X_shard, y_shard, device_W in zip(
X_shards, y_shards, device_params)]
for l in ls: # 反向传播在每个GPU上分别执行
l.backward()
# 将每个GPU的所有梯度相加，并将其广播到所有GPU
with torch.no_grad():
for i in range(len(device_params[0])):
allreduce(
[device_params[c][i].grad for c in range(len(devices))])
# 在每个GPU上分别更新模型参数
for param in device_params:
d2l.sgd(param, lr, X.shape[0]) # 在这里，我们使用全尺寸的小批量
def train(num_gpus, batch_size, lr):
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
devices = [d2l.try_gpu(i) for i in range(num_gpus)]
# 将模型参数复制到num_gpus个GPU
device_params = [get_params(params, d) for d in devices]
num_epochs = 10
animator = d2l.Animator('epoch', 'test acc', xlim=[1, num_epochs])
timer = d2l.Timer()
for epoch in range(num_epochs):
timer.start()
for X, y in train_iter:
# 为单个小批量执行多GPU训练
train_batch(X, y, device_params, devices, lr)
torch.cuda.synchronize()
timer.stop()
# 在GPU0上评估模型
animator.add(epoch + 1, (d2l.evaluate_accuracy_gpu(
lambda x: lenet(x, device_params[0]), test_iter, devices[0]),))
print(f'测试精度：{animator.Y[0][-1]:.2f}，{timer.avg():.1f}秒/轮，'
f'在{str(devices)}')pytorch实现
import torch
from torch import nn
from d2l import torch as d2l
from resnet import Residual
# @save
def resnet18(num_classes, in_channels=1):
"""稍加修改的ResNet-18模型"""
def resnet_block(in_channels, out_channels, num_residuals,
first_block=False):
blk = []
for i in range(num_residuals):
if i == 0 and not first_block:
blk.append(Residual(in_channels, out_channels,
use_1x1conv=True, strides=2))
else:
blk.append(d2l.Residual(out_channels, out_channels))
return nn.Sequential(*blk)
# 该模型使用了更小的卷积核、步长和填充，而且删除了最大汇聚层
net = nn.Sequential(
nn.Conv2d(in_channels, 64, kernel_size=3, stride=1, padding=1),
nn.BatchNorm2d(64),
nn.ReLU())
net.add_module("resnet_block1", resnet_block(
64, 64, 2, first_block=True))
net.add_module("resnet_block2", resnet_block(64, 128, 2))
net.add_module("resnet_block3", resnet_block(128, 256, 2))
net.add_module("resnet_block4", resnet_block(256, 512, 2))
net.add_module("global_avg_pool", nn.AdaptiveAvgPool2d((1, 1)))
net.add_module("fc", nn.Sequential(nn.Flatten(),
nn.Linear(512, num_classes)))
return net
net = resnet18(10)
# 获取GPU列表
devices = d2l.try_all_gpus()
# 我们将在训练代码实现中初始化网络
def train(net, num_gpus, batch_size, lr):
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
devices = [d2l.try_gpu(i) for i in range(num_gpus)]
def init_weights(m):
if type(m) in [nn.Linear, nn.Conv2d]:
nn.init.normal_(m.weight, std=0.01)
net.apply(init_weights)
# 在多个GPU上设置模型
net = nn.DataParallel(net, device_ids=devices)
trainer = torch.optim.SGD(net.parameters(), lr)
loss = nn.CrossEntropyLoss()
timer, num_epochs = d2l.Timer(), 10
animator = d2l.Animator('epoch', 'test acc', xlim=[1, num_epochs])
for epoch in range(num_epochs):
net.train()
timer.start()
for X, y in train_iter:
trainer.zero_grad()
X, y = X.to(devices[0]), y.to(devices[0])
l = loss(net(X), y)
l.backward()
trainer.step()
timer.stop()
animator.add(epoch + 1, (d2l.evaluate_accuracy_gpu(net, test_iter),))
print(f'测试精度：{animator.Y[0][-1]:.2f}，{timer.avg():.1f}秒/轮，'
f'在{str(devices)}')
train(net, num_gpus=1, batch_size=256, lr=0.1)